Professional Documents
Culture Documents
Annals of Epidemiology
journal homepage: www.annalsofepidemiology.org
Brief communication
a r t i c l e i n f o a b s t r a c t
Article history: Purpose: The purpose of the study was to examine whether birth characteristics affect the risk of
Received 15 July 2015 rhabdomyosarcoma (RMS) in children and adolescents younger than 19 years.
Accepted 27 November 2015 Methods: A total of 722 RMS cases diagnosed at the age of 0e19 years during 1988e2011 were identified
from the California Cancer Registry and matched by birth date, sex, and race to 2,888 controls using
California birth records. Conditional logistic regression was used to estimate the risk of RMS associated
Keywords:
with birth weight, gestational age, and size for gestational age.
Pediatric rhabdomyosarcoma
Results: High birth weight (odds ratio [OR]: 1.00; 95% confidence interval [CI]: 0.78e1.29) and large for
Birth weight
Fetal growth
gestational age (LGA; OR: 0.94, 95% CI: 0.72e1.23) were not associated with RMS risk overall. Among
non-Hispanic whites, the ORs were 1.33 for high birth weight (95% CI: 0.94e1.89) and 1.17 for LGA (95%
CI: 0.78e1.75); no indications of association were observed for other racial or ethnic groups (P interaction
<.10). Compared with normal gestational age, preterm (<37 weeks) and post-term (>40 weeks) babies
had 16%e18% lower risks of RMS overall, after adjusting for birth weight.
Conclusions: In the largest study to date, there was an indication of association between high birth
weight, LGA, and increased RMS risk among non-Hispanic white children and adolescents, but not in
other racial or ethnic groups.
Ó 2015 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.annepidem.2015.11.007
1047-2797/Ó 2015 Elsevier Inc. All rights reserved.
Please cite this article in press as: Morimoto LM, et al., Birth weight, fetal growth, and risk of pediatric rhabdomyosarcoma: an updated record
linkage study in California, Annals of Epidemiology (2015), http://dx.doi.org/10.1016/j.annepidem.2015.11.007
2 L.M. Morimoto et al. / Annals of Epidemiology xxx (2015) 1e5
Table 1
Risk of rhabdomyosarcoma associated with birth characteristics among 722 cases and 2888 controls in California (1988e2011)
Characteristic All RMS (n cases ¼ 722, n controls ¼ 2,888) eRMS (n cases ¼ 444, n aRMS (n cases ¼ 197, n P for
controls ¼ 1,776) controls ¼ 788) interactiony
Controls, Cases, OR* (95% CI) Controls, Cases, OR* (95% CI) Controls, Cases, OR* (95% CI)
n (%) n (%) n (%) n (%) n (%) n (%)
Birth weight (g)
Low (<2,500) 164 (6) 48 (7) 1.05 (0.73e1.50) 95 27 0.99 (0.63e1.55) 53 13 0.97 (0.51e1.84) .09
Normal (2,500 2,362 (82) 585 (81) 1.00 (Ref.) 1,466 356 1.00 (Ref.) 635 164 1.00 (Ref.)
e4,000)
High (>4,000) 362 (13) 89 (12) 1.00 (0.78e1.29) 215 61 1.09 (0.81e1.47) 100 20 0.89 (0.55e1.44)
P-trend .88 .61 .72
Gestational age (wk)
Preterm (<37) 278 (10) 67 (9) 0.84 (0.62e1.12) 175 40 0.82 (0.57e1.19) 76 16 0.66 (0.38e1.16) .30
Normal (37e40) 1,300 (45) 355 (49) 1.00 (Ref.) 796 217 1.00 (Ref.) 363 103 1.00 (Ref.)
Overdue (>40) 1,134 (39) 252 (35) 0.82 (0.68e0.98) 701 158 0.84 (0.67e1.05) 300 64 0.70 (0.50e0.97)
Missing 176 (6) 48 (7) 1.01 (0.72e1.41)
P-trend .27 .51 .35
Size for gestational age
Small for gestational 277 (10) 73 (10) 1.00 (0.76e1.32) 177 41 0.91 (0.64e1.30) 78 24 1.25 (0.78e1.98) .22
age
Appropriate for 2,099 (73) 523 (72) 1.00 (Ref.) 1,294 325 1.00 (Ref.) 573 139 1.00 (Ref.)
gestational age
Large for gestational 336 (12) 78 (11) 0.94 (0.72e1.23) 201 49 0.96 (0.69e1.33) 88 20 0.94 (0.58e1.53)
age
Missing 176 (6) 48 (7) 1.10 (0.79e1.53) 0.38
P-trend .60 .84
enrolled from 1982 to 1988 observed statistically significant as- Data collection
sociations between both low (OR: 4.46; 95% CI: 1.41e14.1) and
high (OR: 2.41; 95% CI: 1.09e5.35) birth weight and aRMS, The California birth records included data on sociodemographic,
although no statistically significant associations were seen for the birth, and pregnancy characteristics, including birth weight (in
more common embryonal subtype [5]. grams) and gestation length (in days), birth order, plurality, race
The present analysis is an update to previous analyses of RMS and ethnicity (child, maternal, and paternal), parental age and years
cases conducted in California; it includes diagnoses spanning of education, geographic location at birth, month when prenatal
24 years (1988e2011) and ages to 19 years, more than doubling the care began, and principal source of payment for prenatal care and
sample size of earlier studies [8, 10]. This is the largest study delivery. The CCR data included information on the date of cancer
examining the association between birth weight, measures of fetal diagnosis and tumor characteristics.
growth, and risk of RMS overall and by race and ethnic group and Gestational age was calculated from the date of last menstrual
histologic subtype. period through birth date. Outlier values less than 20 weeks or
greater than 43 weeks were removed (n ¼ 73). Size for gestational
age was calculated according to the methodology described in
Methods Alexander et al. [11]. Briefly, information from the pool of all
available “healthy” controls was used to establish standards strat-
Study population ified by week of gestation, sex, and race. Any child above the 90th or
below the 10th percentile for birth weight in their gestational week,
We conducted a population-based case-control study of 722 sex, and race-specific strata was considered to be large for gesta-
cases of pediatric RMS (International Classification of Childhood tional age (LGA) or small for gestational age (SGA).
Cancer, third edition, recode 91; International Classification of Dis- Cancer histologic type was classified as embryonal (ICD-O-3:
eases for Oncology, third edition (ICD-O-3), codes 8900e8905, 8910, 8910, 8912, 8991), alveolar (8920), or unclassified (8900e8902).
8912, 8920, 8991) and 2,888 matched controls by linking cancer Cancer site was classified according to the ICD-O-3 Surveillance
diagnosis records from the California Cancer Registry (CCR) to birth Epidemiology and End Results (SEER) Site/Histology Validation List
records maintained by the vital statistics unit of the California codes (http://seer.cancer.gov/icd-o-3/sitetype.icdo3.d20150918.
Department of Public Health. Cases were diagnosed from 1988 (the pdf), as connective and soft tissue (ICD-10: C470eC476,
earliest year the CCR data were electronically available) through C478eC479, C490eC496, C498eC499), testis (C620eC6221, C629),
2011 (when the linkage was conducted) and born in or after 1978 orbit (eye; C690eC691, C693, C695eC698), nasopharynx
(the earliest year the California birth data were electronically (C110eC113, C118eC119), and others.
available), and each case was linked to statewide birth records and
matched by year and month of birth, sex, race, and ethnicity (i.e.,
Hispanic vs. non-Hispanic) to four controls from the birth records Statistical analysis
who had not been linked to the CCR. Using first and last name, date
of birth, mother’s name, and social security number (when avail- Statistical analyses used both univariate and multivariate con-
able), 76% of RMS cases were linked to a California birth certificate. ditional logistic regression to calculate ORs and 95% CIs for each
This study was approved by Institutional Review Boards at the measure of fetal growth, including birth weight (<2,500,
University of California, Berkeley and the California Department of 2,500e4,000, >4,000 g), gestational age (premature, <37 weeks;
Public Health. normal, 3740 weeks; overdue, >40 weeks), and size for
L.M. Morimoto et al. / Annals of Epidemiology xxx (2015) 1e5 3
gestational age (SGA, appropriate for gestational age, LGA) for total
interactionx
RMS, and for eRMS and aRMS subtypes separately. Demographic
and birth characteristics, including maternal age and race, birth
P for
.057
.170
.025
order, plurality, and start of prenatal care, were assessed as poten-
tial confounders; only plurality changed risk estimates more than
1.96 (0.62e6.25)
0.89 (0.29e2.73)
1.46 (0.55e3.90)
0.91 (0.48e1.73)
1.57 (0.71e3.51)
0.72 (0.28e1.88)
10%. The final model included only plurality, conditioning on birth
OR* (95% CI)
1.00 (Ref.) year, child’s sex, and child’s race. Analyses were stratified by child’s
1.00 (Ref.)
1.00 (Ref.)
race, age at diagnosis or reference age, birth decade, season of birth,
and cancer site. Local regression smoothed curves were constructed
Asian (n cases ¼ 71, n
.34
.46
.18
to portray the proportion of cases at different birth weight, gesta-
controls ¼ 285)
tional age, and birth weight adjusted for gestational age, separately
Cases/controls
37/160
48/202
7/13
5/19
8/21
20/85
11/28
6/36
Results
interactionz
The 722 RMS cases and 2,888 age, sex, and race and ethnicity-
matched controls were similar in most demographic and birth
P for
.061
.247
.052
0.49 (0.16e1.47)
0.98 (0.46e2.11)
0.94 (0.49e1.78)
1.01 (0.40e2.50)
0.60 (0.25e1.39)
27% alveolar, and 11% unclassified. Half of the cancers were located
1.00 (Ref.)
1.00 (Ref.)
1.00 (Ref.)
.88
.34
Cases/controls
numbers.
Neither high birth weight (>4,000 g; OR: 1.00; 95% CI: 0.78e1.29)
56/220
31/126
50/190
nor LGA (>90%; OR: 0.94, 95% CI: 0.72e1.23) was associated with
11/35
4/31
11/41
23/93
8/27
7/43
.147
.254
.077
Risk of childhood rhabdomyosarcoma associated with birth characteristics, stratified by child’s race (1988e2011)
gestational age (37e40 weeks) (OR: 0.82, 95% CI: 0.68e0.98), even
after adjusting for birth weight. This trend, however, was not linear,
1.01 (0.57e1.80)
0.78 (0.50e1.22)
0.94 (0.60e1.47)
0.76 (0.56e1.01)
1.33 (0.87e2.02)
0.87 (0.55e1.38)
1.00 (Ref.)
1.00 (Ref.)
Hispanic (n cases ¼ 288, n
.12
.16
P for interaction from likelihood-ratio test comparing white versus Hispanic populations.
0.63e1.55) and aRMS (OR: 0.89, 95% CI: 0.55e1.44), although the
controls ¼ 1,149)
31/116
146/520
91/446
34/104
25/119
209/859
suggestion that high birth weight (OR: 1.33, 95% CI: 0.94e1.89, P-
17/60
P for interaction from likelihood-ratio test comparing white versus all races.
trend: .08) and LGA (OR: 1.17, 95% CI: 0.78e1.75, P-trend: .10) were
positively associated with total RMS risk among non-Hispanic
whites, but not other racial or ethnic groups (P-interaction ¼ .06
0.83 (0.42e1.64)
1.33 (0.94e1.89)
0.57 (0.32e1.02)
0.86 (0.65e1.15)
0.69 (0.41e1.15)
1.17 (0.78e1.75)
for birth weight; P-interaction ¼ .03 for SGA). The lack of an asso-
OR* (95% CI)
1.00 (Ref.)
1.00 (Ref.)
1.00 (Ref.)
ethnic group in the analysis, where high birth weight (OR: 0.78, 95%
.08
.74
.10
CI: 0.50e1.22, P-trend: .35) and LGA (OR: 0.87, 95% CI: 0.55e1.38, P-
controls ¼ 1,121)
trend: .16) were unrelated to risk. The effect was even more pro-
Cases/controls
132/470
117/492
20/112
206/813
39/133
All RMS
13/54
16/96
Hispanics and non-Hispanic whites (Fig. 1). For birth weight and
Size for gestational age
Normal (37e40)
P-trend
P-trend
diagnosis (<6 years vs. 6 years), cancer site (connective tissue vs.
y
z
x
*
Fig. 1. Proportion of rhabdomyosarcoma cases across (A) birth weight, (B) gestational age, and (C) birth weight/gestational age.
Discussion adult growth and IGF2 being the primary IGF active in early fetal
development. Genes in the IGF axis, as well as in intersecting
In the largest study of pediatric RMS to date, we found an growth related pathways, are highly polymorphic, and genetic
indication of association between high birth weight, LGA, and variation may play an important role in pediatric RMS
increased RMS risk, particularly eRMS, among non-Hispanic white development.
children and adolescents, but not in other racial or ethnic groups. The information collected on birth certificates, which varies
This result supports the associations observed in previous studies, between states and by year, does not contain the data elements
with analyses of primarily white populations yielding stronger necessary to accurately assess fetal growth patterns, such as
positive associations with birth weight [5, 8] compared to studies maternal weight gain, fundal height, and sonograms throughout
with large Hispanic representations [10]. In addition, we observed pregnancy. It has been suggested that proportion of optimal birth
some evidence that longer or shorter periods of gestation (relative weight (POBW) may be a more suitable measure of appropriateness
to normal length) were inversely associated with RMS risk. of fetal growth from neonatal birth records [17]. Perhaps due to
The association of higher birth weight and RMS risk among only limitations of data available from California birth certificates, the
non-Hispanic whites may reflect differences by race and ethnicity. models developed to predict POBW using the method described in
It may also reflect the inability of birth weight to capture fetal Blair et al. [17] explained only approximately 27% of the variation in
growth rates during etiologically relevant time periods. It is known outcome measure (adjusted R-squared ¼ 0.2723), and we did not
that birth weight [12e14] differs by race and ethnicity, with black, include POBW as an independent predictor in our analyses.
Asian, and Hispanic infants having lower gestational age-adjusted This study presents several strengths, including the large sample
birth weight than non-Hispanic whites. However, an analysis of size, ethnic diversity, and unbiased data collection from cases and
U.S. birth records of Mexican American and non-Hispanic whites controls. There are some limitations, however, that should be
from 1994 to 1996 found that Mexican American infants were considered. First, given the design of the study, our conceptualized
heavier than non-Hispanic white infants between 30 and 37 weeks birth cohort did not include cases who were diagnosed in California
of gestation, whereas non-Hispanic whites were heavier at or after but born elsewhere, but it is possible that some of the controls
42 weeks gestation [15], suggesting that intrauterine growth pat- could have moved from California, been diagnosed with RMS
terns [13] differ by ethnic group. It remains unknown how various outside California, and, therefore, not linked to the CCR. Given the
lifestyle, environmental, and genetic factors contribute to this rarity of RMS, the chance of this is low. In addition, there are no data
variation. to suggest that such out-migration is associated with our exposures
Because RMS is a sarcoma thought to originate from muscle cell of interest (birth weight, gestational age). Second, the accuracy of
progenitor cells, with differentiation into primary and secondary birth certificate data is known to vary widely, with measures
muscle fibers occurring during the embryonal and early and mid- of maternal risk factors and comorbidities and complications of
fetal stage [16], fetal growth during these stages may be more pregnancy, labor and delivery having the lowest agreement in a
relevant than birth weight, which may reflect late (>37 weeks) comparison of birth certificates with medical records [18]. The in-
gestation weight gain after muscles are fully developed. Therefore, dependent predictor and adjustment covariates focused on in our
although the utility of high birth weight as an independent risk report, however, were birth characteristics (birth weight, gesta-
factor for RMS is unclear, a role of fetal growth patterns in the tional age) that have excellent (>90%) agreement [19]. Finally,
etiology of pediatric RMS remains plausible. Insulin-like growth changes in the histologic definition of aRMS over the years [20], the
factors (IGFs) play a major role in fetal growth and development, lack of central adjudication of cancer registry cases and differences
with IGF1 being the principal IGF in late fetal development and in diagnosing institution practices may have led to misclassification
L.M. Morimoto et al. / Annals of Epidemiology xxx (2015) 1e5 5